Artificial Intelligence Roles: Who Does What in AI

Building your own artificial intelligence system can be best compared to a team sport. Many forget that artificial intelligence is not just about a technical exercise, but also about the people involved.
By Claudia Virlanuta • Sep 7, 2021

Building your own artificial intelligence (AI) system can be best compared to a team sport. Many forget that artificial intelligence is not just about the technical exercise, but also about the people involved. Arun Chandrasekaran, Distinguished Vice President Analyst at Gartner, affirms the importance of the people working on an AI initiative in a 2020 “Smarter With Gartner” article. He states that “in many organizations, data scientists are still wearing too many hats due to a dearth of talent across other roles.” Chandrasekaran emphasizes an important point, which is that the people in an AI project must have diversified skills and roles throughout your initiative. For this reason, you may need to recruit new people or, with training and refocusing, you may also be able to fill the core artificial intelligence roles with your current employees.

This article will highlight the different roles in an artificial intelligence team and provide you with the tasks, knowledge, soft skills, background, and importance of each of those roles. I’ll also provide some input on the skills you should look for in your existing employees, as well as the ways you can structure your team to work most effectively. I’ll also share the three characteristics of a successful data science team. So keep on reading!

 

Core Roles and Responsibilities in Artificial Intelligence

The following is a list of the core roles and responsibilities in an ideal artificial intelligence team:

 

1. Data Engineer

Data engineers are responsible for building data infrastructure and managing data storage and use. For this reason, data engineers need to keep an eye on who’s using data and what data they’re using. They should also have the ability to implement software in a production environment as needed.

Your data engineer should be highly knowledgeable on matters of storage, computing solutions, and software engineering. They should know how to build data pipelines for your most important projects, as well as how to do data processing at scale. Moreover, they should also know a little about the data science process. For example, they should know about how data is pulled, used, and analyzed.

In terms of soft skills, data engineers are resourceful people who are ready to find answers to questions independently. They work well under pressure and stressful conditions, and they know how to interact and collaborate with people. Data engineers usually come from backgrounds of computer science, computer engineering, or IT. Sometimes they may have other backgrounds in quantitative fields, but they should have some computer science experience. Your machine learning models are highly dependent on the quality of your data, which is why the data engineer role is so important.

 

2. Data Scientist

Data scientists are in charge of running experiments and building machine learning systems. They are also responsible for pulling, cleaning, and analyzing your data. Given that their work is distinct within the team, they also need to be able to explain what they do, along with the results of their experiments, to the other members of the team.

Data scientists are typically trained in statistics, data analysis, data communication, and machine learning. They are also proficient in programming languages like Python, R, or SQL.

Their soft skills should include the ability to thrive in both a team and independent work settings, as well as their ability to learn and absorb new knowledge. It is important that the data scientists on your team have a background in technology, science, or another quantitative discipline. For example, technological backgrounds can include software engineering or programming. Biostatistics is an example of a more scientific background. Even physics, which is considered a quantitative field, paired with data science knowledge from self-study or from an online team training would be a beneficial tool set for the data scientists on your AI team to have.

Data scientists are also key players on a team because they’re the ones who are able to recognize the potential opportunities within your data. In today’s data-heavy and hypercompetitive world, recognizing the potential of data can be a significant competitive advantage, which is why data scientists are in such high demand. And yet, there’s a great shortage of them. Having a clear idea of what a good data scientist does (and does not do) should help you attract and hire great candidates.

 

3. Data Science Manager

The data science manager is accountable for building the data science team. Their main job role is to find, attract, and recruit data scientists and engineers. Once they’ve hired the right team members, the data science manager then helps the data scientists and engineers establish their goals and priorities for a given project in order to maximize their productivity. A data science manager should also be able to pinpoint the problems within your organization that data science can solve.

Soft skills of an ideal data science manager include an ability to match the right people with the right tasks and to provide supportive communication. Managers should also be able to provide effective communication with less technical stakeholders who are involved in a data science project. A good data science manager is able to prioritize issues and be realistic and reasonable about which ones to address first, and how.

Data science managers have a thorough knowledge of the software and hardware that need to be used in a given project. They also fully understand the job roles of their data scientists and data engineers. Usually, data science managers come from a data science background, combined with some form of management training.

Good managers create high-caliber teams, define goals and successes, and focus on impact, all of which are key in high-achieving organizations.

 

Skills to Look For in Your Team

The following is a list of key qualities to look for as you choose the members of your data science team:

 

1. Strong Math Skills

Your data science team should have solid backgrounds in mathematics and statistics. A strong grasp of these subjects will help them know which algorithms need to be applied for a given problem, along with how to explain the insights behind their results.

 

2. A Desire to Learn

Professionals on a data science team should enjoy the process of learning, as well as the challenges that arise along the way. Because data has many abstract qualities, successfully manipulating it requires constant experimentation.

 

3. Creativity

Plenty of challenges will arise for your team as they work on different projects. A creative mindset is essential in order to face these challenges with new solutions and perspectives each time.

 

4. Perseverance

Successful AI and data science projects can take months to be completed and ready for use. Your employees will need perseverance to stay motivated in the face of many challenges and failures until they can call a project complete.

 

5. Adaptability

The AI field is fast-paced and rapidly developing. To keep up, your employees must stay up-to-date with changing technology and should be able to adapt quickly to apply new methods and concepts to their work.

 

6. Passion

Watch for people who show interest in the types of problems and datasets your company deals with the most. If they’ve previously worked on projects related to your industry, there’s a good chance your project and team might pique their interest.

 

7. Reason and Realism

The members of your artificial intelligence team should be pragmatic enough to avoid the project delays that can be caused by perfectionism. Even though machine learning has many uses and even greater potential, it isn’t the perfect answer for solving every problem in your company.

Should a project fail, make sure to do a post-mortem analysis along with your team to identify what was tried, reasons for failure, and when and if you should revisit the project. Moreover, teach your team members about the importance of timeboxing as a method of project time management. Your data science team must set and meet realistic deadlines for project tasks. Timeboxing can help avoid wasting time and resources to finish a task that is “almost done” but affecting schedule.

 

How to Structure Your Data Science Team

Data science teams can have either an embedded or independent structure. Let’s take a look at an overview of each type:

 

1. Embedded

In this type of data science team, your data scientists will be embedded, or inserted, within other functional teams. For example, data scientists could be embedded within marketing or operations teams. This type of team will promote communication and a faster creation of domain expertise within your company. On the flip side, embedded data scientists may also get lonely without other data science colleagues in their immediate group. An embedded data science team structure may also make it difficult for a data scientist to consult with other data scientists when they’re stuck on a problem. If you lead an embedded team, closely monitoring team morale and designating time for your data scientists to spend time together at work can help you avoid these issues.

 

2. Independent

An independent data science team works as group separate from other functional teams within a company. This type of data science team is usually used in larger organizations. As opposed to embedded data scientists, independent data scientists have an easier time asking questions of their data science colleagues and receiving support. One disadvantage of the independent data science team is that they can often become alienated from the problems of the main branches of a business. If you lead an independent data science team, make sure each person on the team has frequent contact with their business partners and stakeholders.

 

Three Characteristics of an Effective AI Team

The first characteristic of an effective AI team is efficient communication. Communication should be part of the culture within your team, especially if you have an independent data science team. Embedded data scientists will naturally have closer communication with their business partners, because data science projects are often started to address problems experienced by other teams in a company. An independent team needs clear lines of communication, as well as clear communication protocols, both within their team and with their stakeholders. Communication is a priority in effective AI teams because data science problems are successfully solved when both the data scientists and the stakeholders are in agreement.

The second characteristic of an effective AI team is support. Ideally, your data scientists should always have someone to support them with data science questions or troubleshooting throughout their project. As I mentioned, an advantage of independent data science teams is that they typically have more options for coworkers to turn to when looking for advice.

The third characteristic of an effective AI team is empowerment. Your data science team should be fueled by empowerment, because the probability of failure in a data science endeavor is always high. Empower your data scientists to communicate their results, even if they’re not always the great success they were hoping for.

You must remember that whether the data science team you work with is embedded or independent, they need to meet and communicate on a regular basis. It’s also good practice for your team to be in continuous contact with their stakeholders. As a leader, make sure you provide all the support and empowerment needed to keep your team motivated and on the track to success.

You now know a bit more about the main roles in an AI team. In their teams, data engineers, data scientists, and data science managers come together through common connections of math, curiosity, a desire to learn, creativity, perseverance, passion, and pragmatism. Whether you choose to keep your data science team independent or embed it within other teams, remember that communication and the right amount of support and empowerment are the keys to setting your team up for success.

Claudia Virlanuta

CEO | Data Scientist

Claudia Virlanuta

Claudia is a data scientist, consultant and trainer. She is the CEO of Edlitera, a data science and machine learning training and consulting company helping teams and businesses futureproof themselves and turn their data into profits.

Before Edlitera, Claudia taught Computer Science at Harvard, and worked in biotech (Qiagen), marketing tech (ZoomInfo), and ecommerce (Wayfair). Claudia earned her degree in Economics from Yale, with a focus on Statistics and Computer Science.